Science Editor

All happy families may be alike, as Leo Tolstoy famously claimed in Anna Karenina, but all happy drug combinations aren't.

In a paper now available online in Molecular Systems Biology, researchers from Cambridge, Mass.-based CombinatoRx Inc., along with colleagues from Boston University and Columbia University, describe how the type of synergistic effects depends on the relationship of the two pathways that are being targeted.

The paper, first author Joseph Lehar told BioWorld Today, provides "a methodology to understand drug-drug interactions," as well as a more general framework for analyzing how biological systems react to multiple changes that happen at once. And that, as CombinatoRx CEO Alexis Borisy told BioWorld Today, is usually what happens in a natural setting.

In the immune system, for example, success in fending off invaders is "not about one cytokine going up or down," Borisy said. "It's about the code of the correct cocktail" of tens to hundreds of cytokines, and chemokines acting in concert."

So far, CombinatoRx is limiting itself to pursuing two-drug combinations - though Lehar, who is director of computational biology at CombinatoRx, said that one advantage of the methods described in the Molecular Systems Biology paper is that they could be used to predict promising combinations of more than two drugs. "Three- and four- and five-way combinations are in principle very sensible things to do, but the space becomes insanely large very quickly," making chemical or genetic screening infeasible.

The scientists compared different types of combination effects, most notably two different types of synergy. One is potentiation, where the addition of one drug allows a second drug to reach its maximal effectiveness at lower doses than without the additional drug (for example, killing 80 percent of bacteria with only half the usual dose.) The other is boosting, where the addition of one drug actually makes a drug more effective than it could be without the other drug (for example, killing a maximum of 100 percent instead of 80 percent of bacteria.)

In mathematical modeling studies, they found that "these response shapes are telling you something interesting about the mechanism of the combination effect," Lehar said. Such a combination effect is not easily predictable from the effects of the single agents.

Specifically, the results suggest that a boosting response to two drugs made it likely that the two drugs were inhibiting parallel cellular pathways; potentiation tended to occur when the drugs targeted negative feedback loops within a pathway.

The researchers confirmed their computational models by testing combination effects on sterol metabolism in yeast cells. The mechanism of sterol metabolism in yeast has been worked out in great detail and has a number of inhibitors, making it a good system to test mathematical modeling predictions.

When they used different two-inhibitor combinations in yeast, the results "confirmed exactly what we would have expected from our simulations," Lehar said. "It's another validation that what we're pulling out is biology instead of random noise."

For others who are interested in using the same methods on their own research, "all of the algorithms are completely revealed" in the paper, Lehar said. "Any competent programmer could program it very quickly." They will have to, since CombinatoRx has no plans to share their actual programs: "We're not in the software licensing business."